Repository navigation
Conversation
|
Hi @rich7420 |
Owner
Author
|
@kazuyukitanimura Yes, the current version is ready for review at apache/datafusion-comet#6069. This fork PR preserves an older CI snapshot iits changes are included there |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
Related to apache#6043 and apache#6069. Fork CI verification only; this PR does not close an upstream issue.
Rationale for this change
Sliding ANSI/TRY integer SUM needs Spark-compatible overflow detection and recovery while keeping native window execution. This fork draft preserves the verified
ce6e3a90dsnapshot with Spark SQL and benchmark CI coverage. Upstream apache#6069 has since advanced to67889aba5; this older fork no longer matches its current head.What changes are included in this PR?
The head is
ce6e3a90dc0d2333f8cf0c6a8c8b60c53e96b054, the earlier upstream apache#6069 snapshot. The pinned comparison base is mainc7aa38bd0b1e7dde39cf650da43a4452a7a5ea4c.The six-file feature diff contains the native sliding integer accumulator, planner routing, window regression tests, SQL fixtures, benchmark and expression audit. It preserves intermediate overflow and TRY recovery with exact i128 prefix sums and monotonic prefix extrema; update/retract is amortized O(1) per non-null value, and retained capacity is accounted. Legacy integer SUM and ordinary aggregate paths are unchanged. These implementation/test/benchmark changes match the previously verified fork candidate
3e8625af81c6233a4e4b79b82612087ed8b03566; the new head additionally integrates the pinned main.How are these changes tested?
At
3e8625af8, the release native build, 70 Spark 4.1 window tests, four Spark 3.5 strict-warning regressions, ten Rust SUM tests, workspace/all-target release Clippy, formatting and license checks passed locally. The regression includes exhaustive five-value sequences across four frame widths and three evaluation modes. The twelve-case local benchmark (65,536 rows; widths 16/1,024; 0/50/100% NULL; SUM/TRY_SUM) measured native execution 1.29–31.94x faster than Spark within that matrix.The complete earlier fork run passed, including all nine Spark 4.1 SQL shards. A later run at that old head was cancelled and has a red aggregate verdict; it does not establish a code failure or a current complete verdict.
Completed snapshot CI passed at
ce6e3a90d: Required Checks, all Comet suites, Rust, strict Spark 3.5/JDK17 compilation, benchmark checks, TPC-H/TPC-DS and all nine Spark 4.1 SQL shards. Other Spark runtime profiles, Iceberg and macOS were skipped. This remains valid evidence for that snapshot, not for the newer main merge.Upstream apache#6069 now has successful CI at
67889aba5. Fork #30 tracks that current head for additional CI. The cancelled execution job in upstream run36725911344belonged toce6e3a90d; its old maintainer-rerun request is superseded by the new head's successful upstream run.